Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-Supervised Learning for Optical Flow with Generative Adversarial Networks
Wei-Sheng Lai, Jia-Bin Huang, Ming-Hsuan Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add adversarial loss on flow warp error (lambda_adv 0.01) to supervised training (ablation) | supervised SPyNet · KITTI 2012 | EPE: 7.69 -> 7.16 | Improved |
| add adversarial loss (lambda_adv 0.01) (ablation) | supervised SPyNet · KITTI 2015 | EPE: 17.19 -> 16.02 | Improved |
| add adversarial loss (lambda_adv 0.01) (ablation) | supervised SPyNet · Sintel-Final | EPE: 4.70 -> 4.68 | No clear effect |
| increase adversarial weight lambda_adv from 0.01 to 1 (ablation) | proposed model · KITTI 2012 | EPE: 7.16 -> 13.89 | Hurt |
| use discriminator receptive field 47x47 instead of 23x23 (ablation) | PatchGAN discriminator · KITTI 2012 | EPE: 7.38 -> 7.16 | Improved |
| use receptive field 47x47 instead of 95x95 (ablation) | PatchGAN discriminator · KITTI 2012 | EPE: 7.54 -> 7.16 | Improved |
| add classical brightness and smoothness losses on unlabeled KITTI data | supervised SPyNet · KITTI 2012 | EPE: 7.69 -> 10.38 | Hurt |
| use adversarial semi-supervised learning instead of baseline semi-supervised | baseline semi-supervised · KITTI 2015 | Fl: 39.33% -> 38.77% | Improved |
| add unlabeled data with adversarial loss, 25% labeled data | supervised · KITTI 2012 | EPE: 10.60 -> 7.40 | Improved |
| add unlabeled data with adversarial loss, 25% labeled data | supervised · FlyingChairs | EPE: 2.09 -> 2.33 | Hurt |
| add unlabeled data with adversarial loss, 25% labeled data | supervised · Sintel-Clean | EPE: 4.91 -> 3.95 | Improved |
| vary baseline semi-supervised weights lambda_w and lambda_s (ablation) | baseline semi-supervised · KITTI 2012 | EPE: 8.63 to 11.82 | Mixed |
Closest papers by meaning
- Semi-Supervised Learning of Optical Flow by Flow Supervisor (ECCV 2022 · similarity 0.90)
- Unsupervised Deep Learning for Optical Flow Estimation (AAAI 2017 · similarity 0.90)
- Optical Flow Training under Limited Label Budget via Active Learning (ECCV 2022 · similarity 0.90)
- Recurrent Partial Kernel Network for Efficient Optical Flow Estimation (AAAI 2024 · similarity 0.88)
- Taming generative video models for zero-shot optical flow extraction (NeurIPS 2025 · similarity 0.87)
- OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolation (CVPR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
